Limitations of (Procrustes) Alignment in Assessing Multi-Person Human Pose and Shape Estimation

TL;DR

This paper critically examines the limitations of Procrustes alignment in evaluating multi-person 3D human pose and shape estimation, proposing a new method called RotAvat to improve alignment metrics in surveillance scenarios.

Contribution

It introduces RotAvat, a novel technique to refine mesh-ground plane alignment, addressing shortcomings of existing evaluation metrics without realignment.

Findings

RotAvat improves alignment accuracy in qualitative tests

Existing metrics like W-MPJPE and W-PVE have limitations without realignment

RotAvat effectively addresses these limitations in surveillance scenarios

Abstract

We delve into the challenges of accurately estimating 3D human pose and shape in video surveillance scenarios. Beginning with the advocacy for metrics like W-MPJPE and W-PVE, which omit the (Procrustes) realignment step, to improve model evaluation, we then introduce RotAvat. This technique aims to enhance these metrics by refining the alignment of 3D meshes with the ground plane. Through qualitative comparisons, we demonstrate RotAvat's effectiveness in addressing the limitations of existing aproaches.